Computational ghost imaging (CGI) is mainly used to reconstruct grayscale images at present and there are few researches aiming at color images. In this paper, we both theoretically and experimentally demonstrate a colored adaptive compressed imaging method. Benefiting from imaging in YUV color space, the proposed method adequately exploits the sparsity of U, V components in the wavelet domain, the interdependence between luminance and chrominance, and the human visual characteristics. The simulation and experimental results show that our method greatly reduces the measurements required, and offers better image quality compared to recovering red (R), green (G) and blue (B) components separately in RGB color space. As the application of single photodiode increases, our method shows great potential in many fields.
A digital micro mirror device( DMD) is introduced to form Haar wavelet basis , projecting on the color target image by making use of structured illumination, including red, green and blue light. The light intensity signals reflected from the target image are received synchronously by the bucket detector which has no spatial resolution, converted into voltage signals and then transferred into PC .To reach the aim of synchronization, several synchronization processes are added during data acquisition. In the data collection process, according to the wavelet tree structure, the locations of significant coefficients at the finer scale are predicted by comparing the coefficients sampled at the coarsest scale with the threshold. The monochrome grayscale images are obtained under red , green and blue structured illumination by using Haar wavelet inverse transform algorithm, respectively. The color fusion algorithm is carried on the three monochrome grayscale images to obtain the final color image. According to the imaging principle, the experimental demonstration device is assembled. The letter "K" and the X-rite Color Checker Passport are projected and reconstructed as target images, and the final reconstructed color images have good qualities. This article makes use of the method of Haar wavelet reconstruction, reducing the sampling rate considerably. It provides color information without compromising the resolution of the final image.